The infinite sequence a1,a2,… is defined by a1=1 and, for each n≥1, the number an+1 is the smallest positive integer greater than an that has the following property: for each k∈{1,2,…,n}, the number an+1+ak is not a perfect square. Prove that, for all n, it holds that an≤(n−1)2+1.
Problem
Source: Brazil EGMO TST1 2024 #4
Tags: number theory, Sequence, Perfect Squares
13.10.2024 01:34
First terms of this sequence is 1,2,4,6,9,11,13,15. Now suppose that ak≤(k−1)2+1 for k≤N−1 where N≥9. Let's prove that aN≤(N−1)2+1. Consider S={(N−2)2+2,...,(N−1)2+1} and A={a1,...,aN−1}. S and A have no common element. Assume that aN is not in S. |S|=2N−3 and |A|=N−1 hence there must exists some ai such that both (N−2)2+l+ai and (N−2)2+l+m+ai are perfect squares for 2≤l<l+m≤2N−4. √(N−2)2+l+m+ai≥√(N−2)2+l+ai+1(N−2)2+l+m+ai≥(N−2)2+l+ai+1+2√(N−2)2+l+ai2N−4>m≥1+2√(N−2)2+l+ai≥1+2(N−1)≥2N−1Which is impossible as desired.◼
18.01.2025 04:56
Let's use strong induction, base case is trivial. We wish to show an+1≤n2+1. It comes down to a single lemma: Lemma: There is at most one (n−1)2+1<q≤n2+1 for wich ai+q=t2, for each i=1,2,…,n. Proof: If there were q1>q2 we would have: q1−q2=t21−t22≥(n+1)2−n2=2n+1. Because q1,q2>(n−1)2, but that's a contradiction because q1−q2<(n2+1)−((n−1)2+1)=2n−1. ◼ But because there are more than n such q′s (2n−2 to be precise) there must be at least one q for which there is no ai with ai+q=t2 and that means an+1≤n2+1, as we wanted.◼